1. Source Engineer Features
source("EMS.003 Engineer Features.R")
## Parsed with column specification:
## cols(
## timestamp = col_date(format = ""),
## open = col_double(),
## high = col_double(),
## low = col_double(),
## close = col_double(),
## volume = col_double(),
## symbol = col_character()
## )
2. Load Results
results <- read_csv("./Output/results.csv")
## Parsed with column specification:
## cols(
## .default = col_double(),
## timestamp = col_date(format = ""),
## symbol = col_character(),
## future_return_sign = col_integer(),
## close_positive = col_integer(),
## close_negative = col_integer(),
## close_positive_01d = col_integer(),
## close_positive_02d = col_integer(),
## close_positive_03d = col_integer(),
## close_positive_04d = col_integer(),
## close_positive_05d = col_integer(),
## close_positive_06d = col_integer(),
## close_positive_07d = col_integer(),
## close_positive_08d = col_integer(),
## close_positive_09d = col_integer(),
## close_positive_10d = col_integer(),
## close_positive_20d = col_integer(),
## close_positive_30d = col_integer(),
## close_positive_40d = col_integer(),
## close_positive_50d = col_integer(),
## close_positive_60d = col_integer()
## # ... with 3 more columns
## )
## See spec(...) for full column specifications.
2. Examine Prediction
ggplot(results, aes(x = pred)) +
geom_histogram(binwidth = 0.1, fill = "blue", alpha = 0.6) +
geom_vline(xintercept = 0)
3. Table Ranked by Pred
results_top <- results %>%
group_by(symbol) %>%
mutate(pred = round(pred, 2),
pred_7d = round(pred - lag(pred, 5), 2),
pred_1m = round(pred - lag(pred, 21), 2),
close = round(close, 2)) %>%
filter(timestamp == max(timestamp)) %>%
select(symbol, pred, pred_7d, pred_1m, close) %>%
arrange(desc(pred)) %>%
ungroup() %>%
filter(row_number() <= 50)
DT::datatable(results_top)
4. Table Ranked by Pred 7d
results_top <- results %>%
group_by(symbol) %>%
mutate(pred = round(pred, 2),
pred_7d = round(pred - lag(pred, 5), 2),
pred_1m = round(pred - lag(pred, 21), 2),
close = round(close, 2)) %>%
filter(timestamp == max(timestamp)) %>%
select(symbol, pred, pred_7d, pred_1m, close) %>%
arrange(desc(pred_7d)) %>%
ungroup() %>%
filter(row_number() <= 50)
DT::datatable(results_top)
5. Table With Features
results_features <- results %>%
group_by(symbol) %>%
mutate(pred = round(pred, 2),
close_change_10d = round(close_change_10d, 2),
close_drawdown_03m = round(close_drawdown_03m, 2),
close_drawup_03m = round(close_drawup_03m, 2),
close_positive_10d = round(close_positive_10d, 2)) %>%
filter(timestamp == max(timestamp)) %>%
select(symbol, pred, close_change_10d, close_drawdown_03m, close_drawup_03m, close_positive_10d) %>%
arrange(desc(pred)) %>%
filter(row_number() <= 50)
DT::datatable(results_features)
6. Table With Sparklines
results_spark <- results %>%
filter(symbol %in% results_top[["symbol"]]) %>%
group_by(symbol) %>%
summarise(pred = as.character(htmltools::as.tags(sparkline(c(pred)))),
close = as.character(htmltools::as.tags(sparkline(c(close))))) %>%
formattable() %>%
formattable::as.htmlwidget()
results_spark[["dependencies"]] <- c(results_spark[["dependencies"]],
htmlwidgets:::widget_dependencies("sparkline", "sparkline"))
results_spark
7. Plots
ggplot(results %>% filter(symbol == "FB"), aes(x = timestamp)) +
geom_line(aes(y = close, colour = pred), size = 1.5) +
scale_colour_gradient2(low = "red", mid = "gray", high = "blue", midpoint = 0) +
labs(title = "Facebook (FB)", subtitle = "One year chart with prediction overlay")
ggplot(results %>% filter(symbol == "AAPL"), aes(x = timestamp)) +
geom_line(aes(y = close, colour = pred), size = 1.5) +
scale_colour_gradient2(low = "red", high = "blue", mid = "gray", midpoint = 0) +
labs(title = "Apple (AAPL)", subtitle = "One year chart with prediction overlay")
ggplot(results %>% filter(symbol == "NFLX"), aes(x = timestamp)) +
geom_line(aes(y = close, colour = pred), size = 1.5) +
scale_colour_gradient2(low = "red", high = "blue", mid = "gray", midpoint = 0) +
labs(title = "Netflix (NFLX)", subtitle = "One year chart with prediction overlay")
ggplot(results %>% filter(symbol == "GOOG"), aes(x = timestamp)) +
geom_line(aes(y = close, colour = pred), size = 1.5) +
scale_colour_gradient2(low = "red", mid = "gray", high = "blue", midpoint = 0) +
labs(title = "Google (GOOG)", subtitle = "One year chart with prediction overlay")
8. Interactive Plots
8.1 Plotly
p01 <- ggplot(results %>% filter(symbol == "FB"), aes(x = timestamp)) +
geom_line(aes(y = close), colour = "blue", size = 1) +
labs(title = "Facebook (FB)", subtitle = "A Plotly interactive graphic")
plotly::ggplotly(p01)
8.2 dygraphs
p01 <- results %>%
filter(symbol == "FB") %>%
select(timestamp, close) %>%
timetk::tk_xts(date_col = timestamp)
## Warning in tk_xts_.data.frame(data = data, select = select, date_var =
## date_var, : Non-numeric columns being dropped: timestamp
## Using column `timestamp` for date_var.
dygraphs::dygraph(p01, main = "Facebook (FB)")
8.3 Highcharter
p01 <- quantmod::getSymbols("GOOG", auto.assign = FALSE)
highcharter::hchart(p01)
9. Faceted Plot
results %>%
filter(symbol %in% c("FB", "AAPL", "NFLX", "GOOG")) %>%
ggplot(aes(x = timestamp)) +
geom_line(aes(y = close, colour = pred), size = 1.5) +
scale_colour_gradient2(low = "red", mid = "gray", high = "blue", midpoint = 0) +
facet_wrap(~ symbol, scales = "free_y")
results %>%
filter(symbol %in% c("FB", "AAPL", "NFLX", "GOOG")) %>%
ggplot(aes(x = timestamp)) +
geom_line(aes(y = pred, colour = pred), size = 1.5) +
scale_colour_gradient2(low = "red", mid = "gray", high = "blue", midpoint = 0) +
facet_wrap(~ symbol, scales = "free_y")
10. Close With Signal
p01 <- ggplot(results %>% filter(symbol == "FB"), aes(x = timestamp)) +
geom_line(aes(y = close, colour = pred), size = 1.5) +
scale_colour_gradient2(low = "red", mid = "gray", high = "blue", midpoint = 0) +
labs(title = "Facebook (FB)", subtitle = "One year price chart with prediction overlay")
p02 <- ggplot(results %>% filter(symbol == "FB"), aes(x = timestamp)) +
geom_line(aes(y = pred, colour = pred), size = 1.5) +
scale_colour_gradient2(low = "red", mid = "gray", high = "blue", midpoint = 0) +
labs(subtitle = "One year prediction chart")
grid.arrange(p01, p02, ncol = 1)
p01 <- ggplot(results %>% filter(symbol == "AAPL"), aes(x = timestamp)) +
geom_line(aes(y = close, colour = pred), size = 1.5) +
scale_colour_gradient2(low = "red", mid = "gray", high = "blue", midpoint = 0) +
labs(title = "Apple (AAPL)", subtitle = "One year price chart with prediction overlay")
p02 <- ggplot(results %>% filter(symbol == "AAPL"), aes(x = timestamp)) +
geom_line(aes(y = pred, colour = pred), size = 1.5) +
scale_colour_gradient2(low = "red", mid = "gray", high = "blue", midpoint = 0) +
labs(subtitle = "One year prediction chart")
grid.arrange(p01, p02, ncol = 1)
p01 <- ggplot(results %>% filter(symbol == "NFLX"), aes(x = timestamp)) +
geom_line(aes(y = close, colour = pred), size = 1.5) +
scale_colour_gradient2(low = "red", mid = "gray", high = "blue", midpoint = 0) +
labs(title = "Netflix (NFLX)", subtitle = "One year price chart with prediction overlay")
p02 <- ggplot(results %>% filter(symbol == "NFLX"), aes(x = timestamp)) +
geom_line(aes(y = pred, colour = pred), size = 1.5) +
scale_colour_gradient2(low = "red", mid = "gray", high = "blue", midpoint = 0) +
labs(subtitle = "One year prediction chart")
grid.arrange(p01, p02, ncol = 1)
p01 <- ggplot(results %>% filter(symbol == "GOOG"), aes(x = timestamp)) +
geom_line(aes(y = close, colour = pred), size = 1.5) +
scale_colour_gradient2(low = "red", mid = "gray", high = "blue", midpoint = 0) +
labs(title = "Google (GOOG)", subtitle = "One year price chart with prediction overlay")
p02 <- ggplot(results %>% filter(symbol == "GOOG"), aes(x = timestamp)) +
geom_line(aes(y = pred, colour = pred), size = 1.5) +
scale_colour_gradient2(low = "red", mid = "gray", high = "blue", midpoint = 0) +
labs(subtitle = "One year prediction chart")
grid.arrange(p01, p02, ncol = 1)
